Probabilistic Feature Attention as an Alternative to Variables in Phonotactic Learning

Since Halle (1962), explicit algebraic variables (often called alpha notation) have been commonplace in phonological theory. However, Hayes and Wilson (2008) proposed a variable-free model of phonotactic learning, sparking a debate about whether such algebraic representations are necessary to capture human phonological acquisition. While past experimental work has found evidence that suggested a need for variables in models of phonology (Berent et al. 2012, Moreton 2012, Gallagher 2013), this paper presents a novel mechanism, Probabilistic Feature Attention (PFA), that allows a variable-free model of phonotactics to predict a number of these phenomena. Additionally, experimental results involving phonological generalization that cannot be explained by variables are captured by this novel approach. These results cast doubt on whether variables are necessary to capture human-like phonotactic learning and provide a useful alternative to such representations.

Learning Nonlocal Phonotactics in a Strictly Piecewise Probabilistic Phonotactic Model

Phonotactic learning is a crucial aspect of phonological acquisition and has figured significantly in computational research in phonology (Prince & Tesar 2004). However, one persistent challenge for this line of research is inducing non-local co-occurrence patterns (Hayes & Wilson 2008). The current study develops a probabilistic phonotactic model based on the Strictly Piecewise class of subregular languages (Heinz 2010). The model successfully learns both segmental and featural representations, and correctly predicts the acceptabilities of the nonce forms in Quechua (Gouskova & Gallagher 2020).

Modelling Mandarin speakers’ phonotactic knowledge

This paper investigates the nature of native Mandarin Chinese speakers’ phonotactic knowledge via an experimental study and formal modelling of the experimental results. Results from a phonological well-formedness judgement experiment suggest that Mandarin speakers’ phonotactic knowledge is sensitive not only to lexical statistics, but also to grammatical principles such as systematic and accidental phonotactic constraints, allophonic restrictions and segment–tone co-occurrence restrictions. We employ the UCLA Phonotactic Learner to model Mandarin speakers’ phonotactic knowledge, and compare the model's well-formedness predictions with speakers’ judgements. The disparity between the model's predictions and the well-formedness ratings from the experiment indicates that grammatical principles and the lexicon are still not sufficient to explain all of the variations in the speakers’ judgements. We argue that multiple biases, such as naturalness bias, allophony bias and suprasegmental bias, are effective during phonotactic learning.

Exceptionality and derived environment effects: a comparison of Korean and Turkish

Morphologically derived environment effects (MDEEs) are well-known examples where phonotactic patterns in the lexicon mismatch with what is allowed at morphological boundaries – alternations. Analyses of MDEEs usually assume that the alternation is morphologically general, and that the sequences ‘repaired’ across morpheme boundaries are phonotactically well-formed in the lexicon. This paper examines the phonotactic patterns in the lexicon of two languages with MDEEs: Korean palatalisation and Turkish velar deletion. I show that Korean heteromorphemic sequences that undergo palatalisation are underattested in the lexicon. A computational learner learns a markedness constraint that drives palatalisation, suggesting a pattern of exceptional non-undergoing. This contrasts with Turkish, where the relevant constraint motivating velar deletion at the morpheme boundary is unavailable from phonotactic learning, and where the alternation is an example of exceptional triggering. These results indicate that MDEEs are not a unitary phenomenon, highlighting the need to examine these patterns in closer quantitative detail.

Accidental Gaps and Surface-Based Phonotactic Learning: A Case Study of South Bolivian Quechua

The lexicon of a natural language does not contain all of the phonological structures that are grammatical. This presents a fundamental challenge to the learner, who must distinguish linguistically significant restrictions from accidental gaps ( Fischer-Jørgensen 1952 , Halle 1962 , Chomsky and Halle 1965 , Pierrehumbert 1994 , Frisch and Zawaydeh 2001 , Iverson and Salmons 2005 , Gorman 2013 , Hayes and White 2013 ). The severity of the challenge depends on the size of the lexicon ( Pierrehumbert 2001 ), the number of sounds and their frequency distribution ( Sigurd 1968 , Tambovtsev and Martindale 2007 ), and the complexity of the generalizations that learners must entertain ( Pierrehumbert 1994 , Hayes and Wilson 2008 , Kager and Pater 2012 , Jardine and Heinz 2016 ). In this squib, we consider the problem that accidental gaps pose for learning phonotactic grammars stated on a single, surface level of representation. While the monostratal approach to phonology has considerable theoretical and computational appeal ( Ellison 1993 , Bird and Ellison 1994 , Scobbie, Coleman, and Bird 1996 , Burzio 2002 ), little previous research has investigated how purely surface-based phonotactic grammars can be learned from natural lexicons (but cf. Hayes and Wilson 2008 , Hayes and White 2013 ). The empirical basis of our study is the sound pattern of South Bolivian Quechua, with particular focus on the allophonic distribution of high and mid vowels. We show that, in characterizing the vowel distribution, a surface-based analysis must resort to generalizations of greater complexity than are needed in traditional accounts that derive outputs from underlying forms. This exacerbates the learning problem, because complex constraints are more likely to be surface-true by chance (i.e., the structures they prohibit are more likely to be accidentally absent from the lexicon). A comprehensive quantitative analysis of the Quechua lexicon and phonotactic system establishes that many accidental gaps of the relevant complexity level do indeed exist. We propose that, to overcome this problem, surface-based phonotactic models should have two related properties: they should use distinctive features to state constraints at multiple levels of granularity, and they should select constraints of appropriate granularity by statistical comparison of observed and expected frequency distributions. The central idea is that actual gaps typically belong to statistically robust feature-based classes, whereas accidental gaps are more likely to be featurally isolated and to contain independently rare sounds. A maximum-entropy learning model that incorporates these two properties is shown to be effective at distinguishing systematic and accidental gaps in a whole-language phonotactic analysis of Quechua, outperforming minimally different models that lack features or perform nonstatistical induction.